I've been running a handful of MCP servers locally and remotely and I'm curious what others settled on. A few questions: stdio or HTTP transport in production? How do you scope which tools each agent can see? Any good patterns for budgets or rate limits on tool calls? And which servers turned out to be genuinely useful vs. demo-only? I'll share mine in the comments.
Top comments (2)
On scoping: do you filter which tools an agent sees in the client config, or does the server decide per agent? We're the Auten team (we build a computer-use MCP, so read this with that bias), and for anything that touches a real screen we found a call budget matters less than a hard confirm step before buy, delete or send. Looking forward to your list.
Transport: stdio for anything local, HTTP only when the server has to be shared. For scoping, a per-agent allowlist in client config is easier to audit than server-side filtering.